Migration Fest

By Zexyen ā€¢ 3 minutes read ā€¢


Table of Contents

Insanity Ahoy

I have been starting to share more of my self-hosted services with other people, which means I have to implement my own SLA & Uptime insuranceā€¦ Everything was running locally on my Home servers, with Caddy acting as the front-end for all the requests. This worked fine until one of the servers either freaked out, causing a cascading issue of things, or when I had a network outage at home (which is rare). As such I decided to start looking into alternatives.

Introducing the VPS

I have always been interested in using a VPS service, but I wasnā€™t a fan of the pricing / was not fully sure of my own capabilities in configuring it. However, since then I have been provided with a sizable pay increase thanks to my new job, and I am a lot more confident in my skills in running said stack. In terms of Providers, I decided to work with racknerd, as they had the best Price/Perfomance ratio I was looking for.

Migration Plan

At the time, everything was running as so:

My end goal was to have everything unified into one docker system (Komodo), which would allow me to manage updates and additions via Git, simplifying things a lot.

Execution

Moving all of the Komodo stuff in general was fairly simple. I had already stored all the docker data/configs in a dedicated folder on the LXC, so I just had to SFTP it to the new VPS, then configure Komodo. Once that was done most things ā€œJust Workedā€ with the exception of having to Re-ip & re-do a lot of the CORS requests (Particularly for Slink).

Moving the LXCs was similar; with the main issue being with Caddy / PocketIDs Migrations. These were configured using the wonder Proxmox Community Scripts, which while they are a fantastic ā€œOne & Doneā€ solution, they are opinionated and do things in their own ways. In the end I was able to migrate the PocketID DB without any issues, and Caddy was reconfigured with itā€™s own dockerfile to build the custom Caddy Build I already ran on the server.

Problem Childs

When you do a Migration of 4 Core services and ~10 sub-services, you tend to have issues crop upā€¦ My main issue on day 1 was that the original caddy file was referring to everything based on the local IP of the komodo instance on the server (as it was a separate instance). Now that Caddy is on the same Docker network however, I have to redo the entire file and provide it with the actual Ports the services are using, not the translated ports I used originally. After about 2 hours of banging my head on my desk I figured that out and everything came up.

The second problem child was Vaultwarden REALLY not liking the migration, and giving me a ton of issues with reading the DB, mainly because it wasnā€™t properly reading the HTTPS Cert. The aforementioned caddy issue turned out to be the main cause, which after fixing that & re-configuring some other things on the Caddyfile, fixed Vaultwarden.

New Toys

With everything being on a cloud VPS where Iā€™d have to remember a random IP Address to SSH into it, I decided to look into a Server Manager application. Iā€™ve heard a lot about Termix and I decided to configure it and see.